Navigation with Large Language Models: Semantic Guesswork as a Heuristic for Planning (LFG)

Dhruv Shah*, Michael Equi*, Błażej Osiński, Fei Xia, Brian Ichter, Sergey Levine

Berkeley AI Research

Project Page | arXiV | Summary Video

Introduction

This repository contains code used in Navigation with Large Language Models: Semantic Guesswork as a Heuristic for Planning by Dhruv Shah, Michael Equi, Błażej Osiński, Fei Xia, Brian Ichter, and Sergey Levine.

• jupyter_experiment.ipynb - contains the prompts and code used to score the frontiers.
• lfg_demo.ipynb - notebook with examples of using the language scoring to score different frontiers.
• lfg_demo_colab.ipynb - colab version of the above notebook. You can easily run it in your browser!

Installation

To run locally, install the package:

pip install -r requirements.txt

Then simply open lfg_demo.ipynb in jupyter notebook.

LLMs APIs

To use the API locally please provide an API key and organization ID in a .env file in the root directory of the repository.

OPENAI_API_KEY =  sk-***
OPENAI_ORG =  org-***

Citation

If you find this work useful, please consider citing:

@misc{shah2023lfg,
      title={Navigation with Large Language Models: Semantic Guesswork as a Heuristic for Planning}, 
      author={Dhruv Shah and Michael Equi and Blazej Osinski and Fei Xia and Brian Ichter and Sergey Levine},
      year={2023},
      eprint={2310.10103},
      archivePrefix={arXiv},
      primaryClass={cs.RO}
}